Programmable Photonic Solver for Computationally Complex Problems

Computationally complex problems such as the subset sum problem (SSP) are difficult to solve on the electronic computer when the problem size scales up. Optical computing enables the acceleration of complex problems due to its low latency and multiple free degrees of parallelism. Here, we thoroughly exploit the programmability in silicon photonic circuits and construct a programmable microring resonator mesh to implement the solution of two types of computationally complex problems: the SSP and stochastic process simulation. We experimentally accomplish the solving of multiple different sets, which shows the great versatility and scalability of the SSP solver. The knapsack problems are also demonstrated as an example of practical optimization problems derived from SSP. Besides, two-state discrete time stochastic processes are also experimentally accomplished with our solver chip. Our proposed photonic solver shows the potential application of optical NP-complete problem-solving and quantum computing.